While property losses from natural disasters such as Hurricane Matthew may grab headlines in the insurance press, casualty losses often drive the more significant cyclical disruptions in the U.S. market. Fortunately, casualty reserving for recent policy periods looks relatively healthy, although pressure from falling interest rates is starting to mount.
The opening scene of “The Big Short”, a movie adapted from a Michael Lewis book of the same title, closes with Greg Lippman (played by Ryan Gosling) advising the viewer that:
While the whole world was having a big old party, a few outsiders and weirdos saw what no-one else could. These outsiders saw the giant lie at the heart of the economy and they saw it by doing something the rest … never thought to do. They looked.
The next scene then features one of those outsiders, a hedge fund manager named Michael Burry (played by Christian Bale), talking to a potential hire about the “markers” of a bubble in the housing market—ones that had appeared during the housing crash in the 1930’s and that were reappearing in 2005. Mr. Burry was looking at the housing market for signs of a bubble and finding them—well in advance of the housing crash that would follow starting in 2007.
The lesson? One is well-served to look for signs of bubbles before they burst. We last examined the state of the liability market more than 4 years ago (WIN Magazine Summer 2012, Volume 2 Issue 1). It is time for another look.
Like housing bubbles, insurance bubbles also have tell-tale advance markers. The markers are signs that companies have unreasonable expectations about the value of the business that they are putting on the books. Unreasonable expectations sometimes form from necessity. Like homebuyers in the mid-2000’s who were confronted with astronomical housing prices yet rationalized uncomfortable purchase decisions with unreasonable expectations, underwriters in the late 1990’s were faced with the choice of walking away from the market or staying in the game and hoping that the soft market prices weren’t quite as bad as they seemed.
Such false hope and unreasonable expectations can appear in two areas. First, companies can assume that losses are going to be lower than an objective forecast. Second, companies can assume that rates of returns on investments are going to be higher than expected.
Do we see evidence of either delusion in today’s marketplace?
Two useful indicators of the reasonableness of expectations about losses are 1) paid-to-incurred ratios and 2) accident year incurred loss development. As will be shown below, both of these indicators deteriorated significantly in the late 1990’s, well in advance of the market turn in casualty insurance.
At the end of every year, companies must assess the loss experience of the most recent accident year as well as past accident years. In property lines, claims settle quickly, so you usually have a good sense of where you stand at the end of the year. In casualty lines, however, there can be long lags between the occurrence of accidents and the ultimate claims payments. It takes time for suits to be filed and for cases to be settled. It can even take a long time to discover that an injury has even occurred. As a result, early assessments of the total incurred liability loss associated with an accident year will involve a lot of guesswork about how losses will develop in the future.
A good early assessment of this guess work is the paid-to-incurred ratio, which is the ratio of an accident year losses that have actually been paid to the company’s guess of the total incurred loss that will ultimately be paid for claims associated with that accident year. A high value of this ratio suggests that companies have underestimated the ultimate losses.
As can be seen in the figures below, 1 the deterioration in this metric in three of the major lines of casualty insurance was evident in the mid- to late 1990’s. At the time, as the signal emerged, some argued that payment patterns had changed, as industry practices were moving toward accelerated settlement—suggesting that higher paid-to-incurred ratios were not actually a danger sign. It seemed a reasonable story, but, unfortunately, the truth was that payment patterns hadn’t actually changed all that much, and that losses indeed had been significantly underestimated. Reserves were added, eventually triggering the crisis of the early 2000’s.
Paid-to-incurred ratios in recent years, however, look relatively tame. Ratios are well below those seen during the problem years of the 1990’s, are lower than historical averages, and have even been trending downward in recent years.
Quite simply, the idea here is to look at the initial company estimates of the total incurred loss in each accident year and see how they change over time. If companies have to revise their estimates upward, then their initial guesses were too optimistic. Several years in a row of underestimation of the target could suggest a festering problem.
Such a problem emerged in the late 1990’s, as companies were consistently underestimating the losses associated with recent accident years, as can be seen in the Figure 4 below 2
And now? The signals are somewhat mixed, in that Commercial Auto, which led the charge in the late 1990’s has been showing significant upward revisions over the past 5 years. Workers Comp and Other Liability, however, are flat to negative. To be fair, Other Liability did show significant reserve additions at later reports for accident years in the mid-to-late 1990’s. However, these additions were foreshadowed by what were, at the time, unusually high paid-to-incurred ratios—and those are not in evidence here (see Figure 3).
In summary, while one could quibble with certain aspects of this exercise, 3 and the adequacy of loss reserving for recent accident years is worth monitoring going forward, I don’t see a compelling reason to sound an alarm at this point.
Investment Income Expectations
If there are no glaring inadequacies in loss reserving, our attention must shift to the premium side. Specifically, are insurers collecting enough to meet their obligations?
At first glance, pricing metrics seem healthy. Loss ratios to premium, though worse than the peak years of the hard market in the early 2000’s, seem low by historical standards. However, it is here where the ultra-low rate environment can distort our perception. For many of us, our mental benchmarks and rules of thumb for evaluating performance are not automatically adjusted for the fact that invested assets earn much lower rates of return now than in the past.
The decline of interest rates over the decade has affected everyone. On the bright side, when we borrow we may be pleasantly surprised by the terms; rates on fixed rate mortgages, for example, are lower than at any time in recent memory. A dark side, however, emerges when we contemplate our investments and the bleak math of retirement finance. With bond yields at historic lows, many are finding that they must save more and/or work longer than they’d ever anticipated.
A similar principle applies to casualty insurers. Even if they have a good handle on the future obligations associated with policies today, low rates of return mean that they must “save” more than ever before to meet those obligations. In other words, they must collect more premium.
To get an idea of how much this matters, I calculated net loss ratios for each accident year back to 1993, calculated a 1993-2015 average, and compared the latest (2015) loss ratio for each line to its long-term average. I did a similar exercise with present-value “adjusted” net loss ratios where I calculated loss ratios based on the discounted present value of losses using the Treasury curve in each year and the loss payment pattern for each line. 4 I then calculated a 1993-2015 average of that “adjusted” loss ratio and compared it to the 2015 figure in each line. The results are shown in Figure 5.
Net loss ratios today look rather good in historical context when looking at reported values. The 2015 accident year net loss and defense and cost containment ratios were 61% for Other Liability and Workers’ Compensation and 66% for Commercial Auto, all of which are well below long-term averages and thus seem at first glance like spectacular results. However, when we take account of the much lower rates of return on investments that are available today (i.e., the “adjusted” numbers in Figure 5), the results are only about average—a bit better than average for Other Liability, and a bit worse for Workers Comp. Thus, while results as reported seem to compare well to traditional benchmarks, in reality we’re just treading water—we need more premium up front to make up for the fact that we are going to be earning returns on the “float” that are far, far below historical averages.
To wrap up, though I didn’t find anything obviously alarming in the data from recent accident years, I also did not find anything especially comforting. Interest rates continued a decades-long trend by falling throughout most of 2016, and it is too early to tell whether the post-election rise in interest rates marks the end of the trend or just a bump in the road. Companies must accept the reality of low invest returns into the foreseeable future and adjust their expectations and pricing targets accordingly. However, pricing and reserving discipline may be difficult to maintain in a marketplace awash in capital. Stay tuned.
- All figures sourced from SNL Financial. Figures show cumulative net paid divided by total incurred loss and defense and cost containment costs at the latest available evaluation of the first report from Schedule P. Initial evaluations of the first report show a similar story. ↩
- Source is SNL Financial. Development measures change in total net incurred loss and defense and cost containment expenses from first report to second report. ↩
- One issue worth considering is how discounting of certain loss reserves could push paid-to-incurred ratios down over time as interest rates fall. This is likely to be a bigger issue with Workers Comp than with the other lines. ↩
- Specifically, I interpolated annual Treasury yields based on the 1,2,3,5,7,10,20, and 30 year maturities given in the Federal Reserve’s H15 Report and used those interpolated figures as a basis for calculating spot rates. The loss payment pattern was based on the most recent available 10-year average of Schedule P data, with exponential decay assumed to apply to development years after the tenth year. ↩